Automatic classification of patients with mental disorders according to voice dynamics

Abstract

Anomalous aspects of speech and voice, including pitch, fluency, and voice quality, are reported to characterize many mental disorders. However, it has proven difficult to quantify and explain this oddness of speech by employing traditional statistics methods. In this study we employ Recurrence Quantification Analysis (RQA) to investigate the temporal dynamics of voice in three mental disorders. We elicited monological descriptions of short videos in patients with schizophrenia, depression and Asperger's, as well as in related matched controls. We applied RQA to fundamental frequency, speech pause sequences and speech rate. The Rqa indexes (trend and entropy in particular) enable us to quantify and automatically discriminate between populations with >85% of accuracy, highlighting distinctive voice dynamics in each diagnoses.


Back to Table of Contents